Estimated study time: 60 minutes
Content:
Risk management at CFA Level 2 extends beyond identifying risk types to measuring, monitoring, and managing them quantitatively. The primary risk metrics are Value at Risk (VaR), Expected Shortfall (CVaR), scenario analysis, and stress testing. Risk management frameworks distinguish between market risk (losses from changes in market prices), credit risk (losses from counterparty default), liquidity risk (inability to exit positions at fair value), operational risk (failures of people, processes, and systems), and model risk (reliance on models with incorrect assumptions). At Level 2, the focus is on how these risks interact, their measurement methodologies, and how derivatives and other instruments are used to modify the portfolio's risk profile.
Value at Risk (VaR) is the most widely used market risk measure. VaR(X%, T) represents the maximum loss expected over holding period T at confidence level X%. For example, a 1-day 99% VaR of $1M means there is a 1% probability that the portfolio will lose more than $1M in a single day. Three estimation methods are used: (1) Historical simulation — using actual historical return data (past N days) to construct the empirical loss distribution; no distributional assumption required, captures fat tails and correlations implicitly, but limited by the historical sample. (2) Parametric (analytical) VaR — assumes returns are normally distributed: VaR = mu - z*sigma, where z is the appropriate z-score (2.326 for 99%, 1.645 for 95%); computationally efficient but understates tail risk. (3) Monte Carlo simulation — generates a large number of random scenarios based on modeled factor distributions and correlations; flexible but computationally intensive.
Expected Shortfall (CVaR, Conditional Value at Risk) addresses VaR's principal limitation — VaR says nothing about the magnitude of losses beyond the VaR threshold. CVaR = E[Loss | Loss > VaR] — the expected loss given that the loss exceeds the VaR cutoff. CVaR is a coherent risk measure (it satisfies subadditivity — the risk of two combined positions is at most the sum of their individual risks), whereas VaR is not always subadditive. Financial regulators (Basel III/IV) have moved toward CVaR (Expected Shortfall) for market risk capital requirements because it better captures tail risk. The practical challenge with CVaR is that it requires more data and is harder to estimate than VaR, especially in the tail of the distribution.
Stress testing and scenario analysis complement VaR by evaluating portfolio performance under specific adverse conditions rather than relying on probability-weighted average loss distributions. Historical scenarios apply actual historical market movements (2008 financial crisis, 1987 Black Monday, COVID March 2020 crash, 1994 bond market rout) to current portfolio positions. Hypothetical scenarios construct plausible but forward-looking adverse conditions (e.g., simultaneous 30% equity decline, 200 bps rate spike, and credit spread widening of 400 bps). Reverse stress testing identifies what combination of market moves would cause the portfolio or institution to fail, then assesses the probability of those conditions. At Level 2, candidates must evaluate the strengths and limitations of each approach and recognize when stress tests reveal risks not captured by VaR.
Risk-adjusted performance measurement is used to evaluate whether the returns generated justify the risks taken. The Sharpe ratio measures reward per unit of total risk. The Treynor ratio = (Rp - rf) / beta_p measures reward per unit of systematic (market) risk. The Information Ratio = (Rp - Rb) / TE measures reward per unit of active risk (tracking error relative to a benchmark). The Sortino ratio = (Rp - rf) / downside deviation uses only downside (negative) volatility in the denominator — relevant when the return distribution is asymmetric (e.g., option strategies or portfolios with negative skewness). Maximum drawdown (the peak-to-trough decline over a period) measures the worst historical loss experienced, relevant for investors who cannot sustain large cumulative losses.
Key Terms:
Quiz Questions:
Q1. A portfolio has a daily expected return of 0.05% and a daily standard deviation of 1.5%. Assuming normal distribution, calculate the 1-day 99% VaR in dollar terms for a $10M portfolio.
A) VaR = (mu - z*sigma) * Portfolio = (0.05% - 2.326 * 1.5%) * $10M = (0.05% - 3.489%) * $10M = -3.439% * $10M = $343,900 loss. B) VaR = z*sigma * Portfolio = 2.326 * 1.5% * $10M = 3.489% * $10M = $348,900. C) VaR = sigma * Portfolio = 1.5% * $10M = $150,000 (1 standard deviation, 84% confidence only). D) VaR = z * Portfolio = 2.326 * $10M = $23.26M.
Answer: A — Parametric VaR: the loss at the 99th percentile of the daily return distribution. For a one-tailed 99% level, z = 2.326. VaR (as a loss) = -(mu - z*sigma) * Portfolio Value = -(0.05% - 2.326*1.5%) * $10M = -(-3.439%) * $10M = 3.439% * $10M = $343,900. The portfolio is expected to lose more than $343,900 on 1% of trading days (approximately 2-3 times per year). Option B ignores the expected return contribution (which is small over one day but should be included for completeness).
---
Q2. A risk manager computes 1-day 95% VaR using three methods: Historical Simulation = $500,000; Parametric (normal) = $400,000; Monte Carlo = $520,000. Which statement best interprets the difference between historical simulation and parametric VaR?
A) Historical simulation is always more accurate than parametric VaR. B) The historical simulation VaR ($500K) exceeds parametric VaR ($400K), which is consistent with historical returns exhibiting fat tails (excess kurtosis) and negative skewness — extreme losses occurred more frequently in the historical data than the normal distribution assumption implies. C) Parametric VaR is higher because it uses a different confidence level. D) The difference is caused solely by differences in the time period used in each method.
Answer: B — Financial returns are well-documented to have "fat tails" — more extreme losses occur than the normal distribution predicts. Historical simulation captures this empirically because it uses actual return data without imposing a distributional assumption. Parametric VaR (assuming normality) underestimates tail risk, producing a lower VaR estimate. When historical simulation VaR exceeds parametric VaR, it is typically evidence of non-normality (fat tails/negative skewness) in the return distribution. Monte Carlo's higher estimate may reflect the specific model assumptions about tail behavior used in the simulation.
---
Q3. An analyst evaluates two portfolios with the following statistics. Portfolio X: Return = 14%, Standard Deviation = 18%, Beta = 1.2, Tracking Error = 6%. Portfolio Y: Return = 12%, Standard Deviation = 14%, Beta = 0.9, Tracking Error = 3%. Risk-free rate = 3%, Benchmark return = 10%. Calculate the Sharpe ratio, Treynor ratio, and Information Ratio for both portfolios.
A) X: Sharpe = (14-3)/18 = 0.61; Treynor = (14-3)/1.2 = 9.17%; IR = (14-10)/6 = 0.67. Y: Sharpe = (12-3)/14 = 0.64; Treynor = (12-3)/0.9 = 10%; IR = (12-10)/3 = 0.67. B) X dominates Y on all three metrics. C) Y dominates X on all three metrics. D) X has higher IR, Y has higher Sharpe.
Answer: A — Portfolio X: Sharpe = 9/18 = 0.61; Treynor = 9/1.2 = 7.5; IR = 4/6 = 0.67. Portfolio Y: Sharpe = 9/14 = 0.64; Treynor = 9/0.9 = 10.0; IR = 2/3 = 0.67. Portfolio Y outperforms on both Sharpe and Treynor ratios — it generates similar or better risk-adjusted returns with lower risk. Both portfolios have identical information ratios (0.67), meaning both are equally efficient at generating active returns per unit of active risk. The choice between them depends on the investor's benchmark and absolute risk preferences.
---
Q4. A hedge fund manager applies a 2008 financial crisis historical scenario to the current portfolio. The scenario includes: equity markets -40%, credit spreads +600 bps, VIX +30 points, USD strengthening 15%. The fund's equity long positions lose 35%, credit positions lose 12%, and FX positions gain 8% (dollar long). Net portfolio loss = 39%. What does this stress test reveal that VaR may miss?
A) Stress testing confirms that VaR is accurate; both methods give 39% loss. B) The stress test reveals that during the 2008 scenario, correlations between asset classes increased dramatically (equity and credit losses occurred simultaneously, not diversifying each other as in normal periods), and a 39% loss represents a tail event far beyond what VaR (based on normal market correlations) would estimate. C) VaR typically overestimates losses in financial crises. D) Stress testing is redundant when VaR is calculated using historical simulation.
Answer: B — A key VaR limitation is that it typically assumes stable correlations based on normal market history. In financial crises, correlations spike toward 1 as investors sell all risky assets simultaneously — diversification benefits disappear exactly when they are most needed. Stress testing scenarios from historical crises (like 2008) capture these correlation breakdowns and the co-movement of losses across normally uncorrelated assets. This reveals tail risks that VaR (especially parametric VaR with normal distribution assumptions and historical-period correlations) significantly underestimates.
---
Q5. A 60/40 equity/bond portfolio has a maximum drawdown of 35% over its 10-year history (during 2008-2009). A pension fund's board considers this unacceptable — the fund cannot sustain a 35% loss without triggering a forced liquidation at the worst time. Which risk management approach best addresses this specific concern?
A) Reducing the portfolio's Sharpe ratio to decrease volatility. B) Computing a lower VaR by reducing the confidence level. C) Implementing a downside protection strategy — such as a dynamic risk overlay using equity futures (delta hedging portfolio beta down in drawdown scenarios), purchasing put options on the equity index, or establishing a corridor strategy with automated rebalancing triggers that reduce risk exposures when losses approach predefined thresholds. D) Increasing the portfolio's tracking error relative to a 60/40 benchmark.
Answer: C — Maximum drawdown protection specifically addresses path-dependent risk — the risk that a large cumulative loss forces liquidation or breaches funding ratio requirements. Techniques include: dynamic hedging overlays (using futures or options to reduce equity exposure systematically when markets decline), protective puts (capping drawdown at the cost of premium), constant-proportion portfolio insurance (CPPI — maintains a floor by adjusting risky asset exposure), and stop-loss triggers. These strategies directly limit the size of drawdowns, addressing the board's concern in a way that VaR optimization does not. Sharpe ratio optimization (Option A) addresses average risk-adjusted return, not maximum loss.
---